Comparisons of Mercury and Platinum Thermometers. 43 



The corrections to reduce to the Paris hydrogen-scale and 

 the Oallendar-Griffiths' air-scale are respectively 0°"010 and 

 0°-010 at 10°, 0°'004 and a 012 at 20°, and -0°-002 and 

 0°'009 at 30°. Baudin 6166, when its indications are reduced 

 to the absolute scale, reads lower than the Paris nitrogen- 

 scale throughout the range 0° to 31° ; the corrections are 

 0°-030 at 10°, 0°'043 at 20°, and 0°-027 at 30°. The cor- 

 rections to reduce to the Paris hydrogen-scale and the 

 Callendar-Griffiths' air-scale are respectively o, 024 and 

 0°-025 at 10°, 0°-033 and 0°-()41 at 20°, and 0°'016 and 

 0°-027 at 30°. The indications of Baudin 6165, when reduced 

 to the absolute scale, are about o, 035 too low throughout 

 the range 10° to 30°. 



As will be seen from the curves for the " capacity for heat 

 of water " shown in fig. 6, the changes in Rowland's values 

 are small, amounting to a decrease of about 1 part in 850 at 

 10° C, the value at about 18° remaining unchanged, while 

 those at 20° and 25° are respectively increased by about 1 in 

 2100 and 1 in 1400. The variation of the specific heat of 

 water with temperature between 15° and 25°, when Rowland's 

 values are referred to the Paris nitrogen-scale, is practically 

 identical with that given by Griffiths' curve. This suggests 

 at once that an explanation of the differences between the 

 mechanical and electrical determinations of the mechanical 

 equivalent must be sought in the energy-measurements. 



In Rowland's experiments an error in the energy- measure- 

 ments may be due either to an error in the diameter of the 

 torsion-wheel of his calorimeter or to the system of weights 

 employed. Inasmuch as the diameter of the torsion-wheel 

 was measured many times by comparison with two standard 

 metre-bars, each of which had been compared with the Coast 

 Survey and other standards, the possibility of an error 

 sufficient to account for the observed differences between the 

 mechanical and electrical determinations must be sought else- 

 where. Turning to the question of the weights used in these 

 experiments, we see that it is not necessary that they be cor- 

 rect absolutely with the standards, if they are only relatively 

 correct, as the formula for the mechanical equivalent contains 

 a weight in both numerator and denominator. On the other 

 hand, if they were not correct relatively, we should hardly 

 expect to find the almost constant difference between the 

 determinations of Griffiths and those of Rowland (reduced 

 to nitrogen-scale) throughout the range 15° to 25°. The 

 evidence accumulated thus far would, we believe, suggest, as 

 a possible explanation of these differences, a still undiscovered 



